feat: add Tiny-NLA activation explanation with trained model weights

9d0d4e9